learning_rate: 5e-06
train_batch_size: 16
eval_batch_size: 8
seed: 42
optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
lr_scheduler_type: linear
lr_scheduler_warmup_steps: 500
training_steps: 5000
mixed_precision_training: Native AMP

Training Loss	Epoch	Step	Validation Loss	Wer
0.4110	0.2762	500	0.9127	38.8028
0.3250	0.5525	1000	0.9034	37.9030
0.2342	0.8287	1500	0.8987	38.5029
0.1750	1.1050	2000	0.9383	38.7653
0.1820	1.3812	2500	0.9534	38.7778
0.1613	1.6575	3000	0.9481	38.7778
0.1852	1.9337	3500	0.9435	38.2279
0.0751	2.2099	4000	0.9913	38.5904
0.0620	2.4862	4500	0.9959	38.8028
0.0814	2.7624	5000	0.9974	38.8903

mogu98/wosota-asr-medium-v2

wosota-asr-medium-v2